INDEX

Explanations

matrix and arguments

The neuron fires on structural or formatting tokens in technical/code documentation—things like LaTeX matrix commands (e.g. “\begin”, “bmatrix”, “left[”), SQL/DDL keywords (“TABLE”, “ROW”), and section‐heading labels (“arguments”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 have

-1.55

woners

-1.43

 would

-1.41

が使

-1.38

 vanaf

-1.34

beforeAll

-1.34

 mogelijk

-1.32

would

-1.28

}());

-1.28

oplasmic

-1.24

POSITIVE LOGITS

for

1.66

それぞれ

1.34

 every

1.33

彼らの

1.29

Their

1.28

 each

1.25

 health

1.19

 thier

1.19

 leurs

1.18

confirmPassword

1.18

Activations Density 0.000%